We found significant, direct associations between migraine headache, as well as its frequency and financial hardship and hospitalizations (for reasons other than childbirth) in the year before the ELSA-Brasil baseline assessment after multivariate adjustment. Also, analyses of individuals in the highest migraine frequency stratum (more than once per week) showed a significant association between migraine and the death of a close relative in the past year. In addition, further adjustment for the report of religious activity and major depression did not significantly change the results.
Although the link between stressors and negative health outcomes is not completely understood, some authors have proposed mechanisms to explain this association. One possible explanation is that the influence of such events on physical health is mediated by the development of depressive and other psychiatric symptoms [3, 35]. However, the presence of major depressive disorder did not influence this association, suggesting that other mechanisms of association may be present. For instance, other authors have highlighted biological pathways to explain the connection between NLEs and physical health. Wittstein et al.  reported a series of cases of cardiac dysfunction precipitated by acute emotional stress, known as the “broken heart” syndrome. Those authors suggested that this syndrome could be related to acute sympathetic stimuli, a neuromodulation pathway also potentially involved in migraine pathophysiology . Similarly, the activation of the hypothalamus-pituitary-adrenal axis in stressful situations [38, 39] may also be associated with a higher migraine activity . In addition, there is some evidence that migraineurs may have stronger and more sustained responses after emotional distress [41, 42], which may predispose these individuals to more deleterious effects caused by stressful situations. This is compatible to the recent concept of allostatic load, in which body responses targeted to compensate and adapt to new situations (as the occurrence of stressors) may accumulate overtime and eventually result in a maladaptive new steady-state . On the other hand, a recent study by Lipton et al.  suggests that fluctuations in stress levels, rather than the absolute perceived stress levels (measured by daily report of the Perceived Stress Scale and the Self-Reported Stress Scale scores) may be an episode trigger. Unfortunately, we cannot explore further this temporality with our data, as it would require a daily record of migraine activity and perceived stress.
In our study, some life events were associated with migraine, whereas others were not. Financial hardship was the NLE with the strongest association with migraine activity in our study, even in full-adjusted models that included family income and depression diagnosis. Lynch et al.  also studied the association between financial hardship and health. Analyzing income data from 1,124 individuals between 1965 and 1983 and physical and psychological functional outcomes 11 years later, they found an increased frequency of depressive symptoms and reduced performance in daily living activities in those who reported more often financial difficulties during that period.
In addition to an association mediated by a higher level of stress caused by financial hardship, and thus higher migraine activity, financial hardship driven by loss of productivity, absenteeism and money spent on medicine because of a higher level of migraine activity may also be mechanisms underlying the association. However, characteristics of our cohort make this reverse causation less probable. The ELSA-Brasil is a multicenter cohort of formally employed civil servants, who are protected by Brazilian laws of the labor force. This is similar to a tenure track, which prevents against income loss because of health-related absenteeism. Most of them are not subject to productivity rules and earn a fixed monthly salary. Yet, some, although not all, medications used for migraine treatment, including prophylaxis, are available free of charge in the country, upon medical prescription. Finding a positive association in this setting is a strong argument that the predominant mechanism of association between financial hardship and migraine are not because of reverse causation and, rather, may be related to higher stress levels.
Hospitalizations for reasons other than childbirth were also associated with migraine in our study, and this association was stronger with higher migraine activity. Although some may advocate that headache attacks could motivate a higher number of hospitalizations, this explanation seems unlikely. Hospitalization because of migraine is not very common. Hawkins et al., studying paid claims for migraine treatment for a total of 215,209 US employees with migraine diagnosis in 2004, found that only a very small amount of the direct related health costs for this condition were relate to inpatient care .
We also found a significant, positive association between death of a close relative and the highest migraine frequency stratum. Also, increasing migraine frequency led to increasing point OR estimates for this association, although this was non-significant in the first two migraine frequency groups. Although our data do not allow to make undoubted affirmatives, it seems reasonable that some positive association exists between those conditions. During grief, the occurrence of somatic symptoms, including headaches, has been recognized for decades [46, 47]. Surprisingly, with the exception of small studies , the association between the loss of a family member and migraine activity is poorly studied.
Robbery and the end of a love relationship had no clear association with migraine activity. Circumstances that motivate financial hardship , hospitalizations  and the consequences of the death of a close relative  may be, in general, more perennial. Although studies about the effect of specific life events on migraine are scarce, our findings are mostly consistent with a recent study by Wardenaar et al. , evaluating the relationship between the occurrence of life events and psychiatric symptomatology. Those authors found that general distress was independently associated to the report of financial problems and illness/injury/victimization. Robbery and the death of a close relative were not associated to general distress in their study. In addition, as pointed out earlier, more recent work on allostatic load includes a new variable to this model, and reinforce the importance of more perennial causes of stress. We may speculate that stress-mediated changes caused by such events may induce a more sustained body response and be related to higher migraine activity. These body changes, according to the allostatic load theory , may trigger a maladaptive cycle and persist even after the cessation of the stressor.
There is another possible explanation for the differences observed for the associations among the studied NLEs and migraine. The influence of causes of persistent stress are more likely to be identified in a cross-sectional study evaluating migraine activity during a 12-month period. We cannot exclude the hypothesis that robbery and the end of a love relationship may have led to a transient increase in migraine activity that was not sufficient to alter the participants’ perception for the whole 12-month period.
We found, for financial hardship, hospitalization other than for childbirth and death of a close relative, increasing odds ratios associated with higher migraine activity. For all of them, the highest migraine frequency stratum (> once per week) was associated with the highest odds ratios. This is consistent to the findings from two other studies. Scher et al.  analyzed individuals with CDH and episodic headache, and found that the occurrence of major life events were associated with headache chronicity. However, in that study, when the analyses were restricted to those with migraine, there was a trend towards a positive association, but with borderline significance (OR 1.15; p = 0.059). We may speculate, from the results of their study and ours, that the occurrence of negative life events may increase migraine activity. Larsson et al.  analyzed predictors of frequent headache (≥ 1 episode per week) in 2,355 teenagers during a one-year follow-up. Those authors found that, in a model including the reduction of leisure time activities, levels of depressive symptoms and gender as independent variables, all of them were significant predictors of frequent headache at follow-up.
The effects of religious attendance on counteracting putative deleterious effects on health outcomes mediated by the occurrence of NLEs have been studied by others, with conflicting results. Bradshaw and Ellison  found religious attendance to attenuate the effects of objective and subjective financial hardship on psychological distress. On the other hand, Kidwai et al. , studying the effect of religious attendance on the association between NLEs and psychological distress, did not find significant influences. In our study, we did not find such effects either, considering any specific life event nor all together. Although this is still a poorly understood field, these conflicting findings may be due to variations in study settings. For example, social support may moderate the effects of religion. There are variations on how these religion-based net supports act across different communities, and consequently, their influences on individual health are heterogeneous . It is important to emphasize that it was beyond the scope of our study to investigate other social and leisure activities related to stress management. We do not draw conclusions from the present work for about the effects of stress management techniques as coping strategies for the occurrence of NLEs nor their relationship with migraine.
This study has some limitations. Because we are analyzing cross-sectional data (as we only have baseline ELSA data at the present moment), it is not possible to infer causality. However, we could adequately study the association between migraine and NLEs, and given the cohort characteristics, the finding of a positive association between migraine and financial hardship is a strong argument toward a stress-related increase in the number of migraine attacks. The questionnaires required individuals to characterize headache episodes and NLEs in the 12 months before the interview, which could have led to recall bias. Our migraine frequency question do not allow to analyze, in separate, those individuals with chronic migraine, (defined by the IHS criteria as more than 15 days/month). As stated earlier, more subtle, transient increases in migraine activity caused by experiencing robbery, death of a relative or the end of a love relationship could remain unidentified by the approach in this study.